Lesson 10: Where are the Limits?

Topics Covered

"Don't bet against AI." History is full of confident assertions about what AI will never do, only to be proven wrong years later.

1. The Context: DIKW Pyramid

To understand intelligence, we must distinguish between levels of understanding:

Data: Raw facts (e.g., 10, 6, 42). A database.
Information: Data with context (e.g., "These numbers are ages of people in a room"). Application software.
Knowledge: Interpretation of information (e.g., "Most people here are under 21"). This is where AI excels today.
Wisdom: Applied knowledge and judgment (e.g., "We should play age-appropriate games"). This remains the frontier.

Many capabilities once considered "impossible" are now standard:

Reasoning: In 1997, IBM Deep Blue beat grandmaster Gary Kasparov at Chess, proving machines could solve complex logical problems.
Natural Language: Systems like Eliza (1965) and IBM Watson (2011, Jeopardy!) paved the way. Today's LLMs understand idiom, nuance, and humor in ways previously thought uniquely human.
Creativity: Generative AI creates novel art and music. Like humans, it "creates" by synthesizing influences from vast datasets of existing work.
Perception: Robots and self-driving cars now perceive and navigate the physical world in real-time.

While we have made massive strides, significant hurdles remain:

AGI (Artificial General Intelligence): We have "narrow" super-intelligence (great at Chess or Protein folding), but not a single system that equals human performance across all domains.
Sustainability: Current models are energy-intensive. Scaling simply by adding more processors is not sustainable; we need smarter, smaller, purpose-built models.
Hallucinations: Generative models can confidently assert falsehoods. Techniques like RAG and verifiers are mitigating this, but it is not fully solved.
EQ & Emotion: AI can simulate emotional intelligence (detecting mood in text), but does it feel joy or sadness? Currently, it lacks deep emotional reciprocity.
Judgment & Wisdom: Ethical decisions, subjective taste (what makes a song a "hit"?), and "common sense" remain difficult to program.

How should humans and AI work together?

Role	Responsibility
Humans	The "What" and "Why". Setting macro goals, defining purpose, ethical judgment, and determining meaning.
AI	The "How". Executing tasks, automating processes, and optimizing workflows to achieve the defined goals.

We are at an inflection point. While limitations exist today, the trajectory suggests we will continue to solve "impossible" problems.